The invention relates to the field of video signal processing, and in particular to the field of interpolating a pixel of an interline of a first field from a sequence of interlaced fields.
Television signals normally contain information for temporally successive fields which can be displayed on the screen, where two successive fields are interlaced (i.e., have different raster positions). Each field contains image information only in every other scan line, where a scan line for which image information is present in the one field is always empty in the second field. In the following discussion, image information values are understood to mean, specifically, brightness values (luminance values) or color values (chrominance values).
FIG. 1 illustrates two such successive fields A and B, where the first field A has scan lines A1, A2, A3, A4 and A5 for which a received video signal contains image information. The field B includes scan lines B1, B2, B3, B4 and B5 for which a received video signal contains image information. The raster positions of the fields A and B are mutually offset in such a way that the positions of the scan lines of the field B correspond to the positions of the empty interlines of the field A, and vice versa.
In order to enhance image quality, it is often desirable to generate complete frames, i.e., images in which no empty interlines are present.
In one particularly simple method, the approach is to double each of the fields, that is, simply to output the content of a superior or inferior scan line as an interline of a field. In this method, continuous brightness or color transitions in the resulting image have a stepped appearance. In addition, edges running diagonally through the image appear stepped or stair-like.
Another known approach to interpolating a pixel in an interline of a field is to utilize the image information values for pixels that lie vertically adjacent to the pixel to be interpolated, and to implement a linear averaging of these image information values. However, in this method, edges running horizontally through the image are not reproduced sharply.
In addition to the pixels of the field which surrounds the pixel to be interpolated, it is also possible to employ the image information values of temporally preceding or following fields for purposes of interpolation.
European Patent EP 0192292 B1 proposes an approach for interpolating a pixel of an interline from a field in which two pixels are selected from the field in which the pixel to be interpolated is located. That is, one pixel from a scan line above the interpolation pixel and one pixel from a scan line below an interpolation pixel, where the horizontal positions of the selected pixels correspond to the horizontal position of the interpolation pixel. In addition, a pixel of a scan line is selected from a temporally following field, where the position of this pixel corresponds to the pixel position of the interpolation pixel. In this known method, the average value of the image information values for these three pixels is determined, and image information of that pixel is selected for which the image information most closely approximates the average value determined. The pixel of the second field is, in other words, utilized for interpolation if the image information of this pixel lies between the image information values for the pixels of the field with the intermediate spot to be interpolated. Otherwise, the image information of one of the pixels from lines adjacent to the interpolation pixel is selected—with the result that one of these values is doubled, i.e., reproduced twice one below the other. However, this mere repetition of a pixel results in a positionally incorrect display of highly vertical transitions, a phenomenon which is particularly noticeable in edges running diagonally in the image.
The publication Kovacevic J; Safranek, R. J.; Yeh, E. M.: Deinterlacing by Successive Approximation, IEEE Transactions on Image Processing, February 1997, pp. 339-344 discloses a method of interpolating a pixel from an interline in which the value of the interpolation pixel is determined utilizing weighting factors, from the pixels lying vertically above and below the interpolation pixel, and from an additional pixel which is selected from a preceding or following field.
U.S. Pat. No. 5,661,525 similarly describes a method of interpolating a pixel in which pixels directly vertically adjacent to the interpolation pixel as well as pixels from preceding or following fields are selected, wherein these selected pixels are weighted to generate the interpolation pixel.
European Patent application EP 0 227 190 describes a method of interpolating a pixel within a field, wherein pixels lying above and below the interpolation pixel are selected and compared. If the difference between these two values is greater than a predetermined value, the larger or the smaller pixel value is selected as the interpolated pixel value. If the difference is smaller than the reference value, one of the two adjacent pixel values, or the average of these pixel values, is selected as the interpolated value.
U.S. Pat. No. 5,910,820 describes a method of interpolating a pixel in which a weighting of selected pixels is implemented depending on a threshold value and depending on the difference between pixel values for the pixels adjacent to the interpolation pixel.
European Patent application EP 0 946 054 A1 describes a method of interpolating a pixel using a weighted 3-tap median filter. The pixels vertically adjacent to the interpolation pixel and the pixels of a temporally preceding or following field are fed to the median filter.
German Patent DE 689 27 957 T2 describes a method of sampling conversion of a television signal, wherein two data values of a first field and a data value of a second field are selected. An average value is generated from the data values of the first field, and a weighted average value is generated from all three data values. One of the three selected data values, the average value of the data values from the first field, or the weighted average value of all data values, is outputted as the interpolated value.
U.S. Pat. No. 5,475,438 describes a method of pixel interpolation in which the average of the pixel values of two pixels is generated which lie vertically adjacent to the interpolation pixel, and in which a pixel is selected from the field preceding the field with the interpolation pixel and from a field following the field with the interpolation pixel. The average value together with the pixel values from the preceding or following field are fed to a median filter, the output value of which is weighted and added to the similarly weighted average value.
There is a need for a system and method for interpolating a pixel of an interline of a field from a sequence of interlaced fields, which ensures a correct display of edges running horizontally through the image, suppression of stepping in diagonally running edges, and a correct reproduction of movements.